Moonshot AI announced the Kimi K3, a powerful AI model with 2.8 trillion parameters. This post discusses its benchmarking performance against competitors and reflects on the implications of these models through a humorous benchmark involving a pelican riding a bicycle. The author provides insights into the costs and efficiencies of Kimi K3 compared to other models and discusses the value derived from running simple prompts to explore model characteristics. Despite being a playful experiment, the author underscores the importance of understanding how these models function and their capabilities in practical applications.